Laboratory dataset for a locally fabricated flat Coanda-effect screen: 52 clear-water runs and model capacity predictions

Abstract

Laboratory dataset for a locally fabricated flat Coanda-effect screen Fifty-two clear-water runs on a flat Coanda-effect screen intake fabricated from locally available materials, with the measured capture compared against the predictions of the Coanda-effect screen analysis software distributed by the US Bureau of Reclamation. The screen panel carries 32 wedge wires machined from 8 mm square steel bar: wire width 7.6 mm, slot opening 0.8 mm, tilt angle 4.75°, effective width 0.18 m, length along the flow 0.273 m. It was tested in a 0.20 m glass-walled flume behind four interchangeable Ogee accelerator plates combining drop heights of 0.10 m and 0.15 m with screen tangency angles of 45° and 50°, over inflows from 2.24 L/s to 21.36 L/s. The inflow was determined from the head over the accelerator-plate crest and the captured discharge was gauged with a 60° V-notch thin-plate weir. The campaign was carried out in November 2014. Files coanda_screen_dataset_52runs.csv — all 52 runs Column Units Meaning configuration — accelerator-plate case, A to D plate_drop_height_m m drop height of the plate, hp tangency_angle_deg ° tangency angle of the screen panel, θ head_ogee_crest_m m measured head over the crest head_vnotch_m m measured head over the V-notch Q_in_crest_Ls L/s inflow from the crest rating relation Q_R_weir_Ls L/s captured discharge measured at the weir Q_T_model_Ls L/s capture predicted by the model for the same inflow and geometry Q_bypass_model_Ls L/s bypass discharge reported by the model deviation_Ls L/s QT − QR, positive for overprediction ratio_QR_QT_pct % 100 QR / QT informative 0/1 1 if the run tests the capture model, 0 if the model returns full capture model_reruns_results.csv — the 34 informative runs, recomputed Predictions under two discharge relations and two inflow-width conventions, together with the flow state the model computes at the leading edge of the screen. Column Units Meaning config, hp_m, theta_deg, Qin_Ls, QR_measured_Ls — as above QT_2013_w018_Ls, QT_2013_wcorr_Ls L/s angle-of-attack relation of Wahl (2013), inflow over 0.18 m and over 0.20 m QT_2021_w018_Ls, QT_2021_wcorr_Ls L/s offset-based relation of Wahl et al. (2021), same two conventions V0_ms, d0_mm m/s, mm velocity and depth at the leading edge Froude, Reynolds, Weber — dimensionless groups at the leading edge delta_psi_deg ° angle of attack of the flow to the slot wetted_len_m m wetted length of the panel slope_statistics.py Regression and interaction analysis of the growth of the relative deviation with inflow, including the test of the difference between the 45° and 50° tangency angles. Reproduces the statistics reported in the paper from model_reruns_results.csv. Notes on precision and on the informative flag Q_in_crest_Ls and the measured heads carry three decimals; Q_R_weir_Ls is recorded to two decimals; Q_T_model_Ls and Q_bypass_model_Ls are the model output at its own reporting resolution of 0.1 L/s. Because of that, the identity Qin = QT + Qbypass holds only to within about 0.1 L/s, and a reported bypass of 0.0 does not by itself mean that the model predicted full capture. The informative flag therefore follows the predicted capture rather than the reported bypass: a run is informative when |Qin − QT| > 0.01 L/s, that is, when the model predicts that some of the inflow leaves the panel. Thirty-four of the 52 runs meet that criterion and are the basis of every agreement metric in the paper; in the remaining eighteen the model returns full capture, so predicted and measured capture coincide by mass conservation whatever the capture model does, and those runs test the gauging chain instead. A related but distinct count appears in the paper: nineteen runs captured the entire inflow physically, that is QR = Qin. These are the runs used to cross-check the two gauging devices against each other. The two sets differ by one run, case B at 6.222 L/s, where the screen captured everything while the model predicts a bypass of about 0.02 L/s; that run is used in the cross-check and is also one of the 34 informative runs. Software The Coanda-effect screen analysis software under assessment was written by T. L. Wahl of the US Bureau of Reclamation, is in the public domain, and is distributed by the Bureau. It is not redistributed in this deposit. Its recommended citations are: Wahl, T.L. Hydraulic performance of Coanda-effect screens. Journal of Hydraulic Engineering 2001, 127(6), 480–488. Wahl, T.L. New testing of Coanda-effect screen capacities. HydroVision International 2013, Denver, CO, 23–26 July 2013. License The data files and the analysis script are released under CC BY 4.0.